Abacum launches AI agent for finance model building

Abacum has launched Modeler, an AI agent for financial planning and analysis teams. The product is available to all Abacum customers as part of the company's existing platform.

Modeler is designed to let finance teams create business models from natural-language prompts or existing spreadsheets. The tool builds live models on connected company data and presents the proposed structure and logic for approval before anything is saved.

The launch expands Abacum's push into what it calls agentic modelling, in which finance teams define what a model should do and the software builds it. Instead of constructing models line by line, users can request a profit and loss statement from general ledger data, upload an Excel file for reconstruction on live data, or ask for changes to an existing model.

The product uses established model blueprints, maps them to connected datasets, and applies calculation logic through Abacum's existing engine. It also includes an explanation feature that lets users ask how a model works and receive a line-by-line account of variables, links, and calculations.

Finance bottleneck

Abacum targets a long-standing issue for finance teams: a small number of specialists often build and maintain complex planning models. That can create delays when those staff are unavailable and leave companies exposed when knowledge of how models work is concentrated in one person or a small group.

The new product is intended to reduce the time spent building and maintaining models, shifting effort toward analysis, scenario planning, and decision-making. Tasks that often take days or weeks can be shortened to minutes, while finance teams retain formal control through an approval step before a model is adopted.

The system is also positioned as a way to move older spreadsheet-based work into a shared environment. Users can upload an Excel file, and the software reads lines, sections, and totals before rebuilding the model using data already connected to the Abacum platform.

Company push

Abacum, founded in 2020, sells planning software to finance teams and says it serves more than 500 customers in over 40 countries. Named customers include DISH, ThoughtSpot, Strava, Acorns, and Replit.

The company is headquartered in New York and also has offices in Toronto, Barcelona, Buenos Aires, and the UK. It says it has raised more than USD $100 million from investors including Scale Venture Partners, Atomico, and Y Combinator.

Julio Martínez, Co-Founder and Chief Executive Officer of Abacum, said the product responds to pressure on finance teams to deliver more output with limited time and headcount.

"Finance has been told to do more with less for as long as anyone can remember. We think that promise has run its course. Modeler is about doing more with more: more models, built in minutes instead of weeks, by more of the team, with less lift than it has ever taken. The agent does the building, Abacum's Engine still does the math, and finance still approves every model. What changes is how fast you get there, and who gets to build," said Martínez, Co-Founder and Chief Executive Officer, Abacum.

Model transparency is a central part of the pitch. Users can review every proposed variable and the reasoning behind it before approving a model, and teams can ask the software to explain the logic of an existing model without manually tracing formulas or dependencies.

That feature is aimed in part at onboarding and resilience. In many finance departments, new staff must work backwards through inherited spreadsheets and models to understand assumptions, formula chains, and version changes, a process that can be slow and prone to error.

Martínez said the company sees that issue as broader than a software design problem.

"Every finance leader knows the model that lives in one person's head. That isn't a modeling problem, it's a resilience problem. When more of the team can ask for a model, see exactly how it works, refine it with natural language, and ask it to explain itself, the model belongs to the business. That's what AI-native should mean: not AI bolted onto the side of a platform, but a finance team that is AI-native itself," said Martínez.